Mimicking postseismic creep in the laboratory: Testing models for transient creep in the upper mantle

Predictions of postseismic creep, glacial isostatic adjustment (GIA), and seismic-wave attenuation rely on a sound understanding of the microphysics of transient rheological behavior of olivine-rich rocks, the main constituent of the upper mantle. Recent work proposes that changes in dislocation density and dislocation interactions in olivine may explain the time-dependent evolution of the viscosity of the upper mantle as inferred from geodetic studies. We designed load-relaxation experiments to test whether this model (also known as the backstress model) can accurately predict the transient rheological behavior of polycrystalline olivine during load relaxations similar to those experienced by the upper mantle during postseismic creep and GIA.

We performed our experiments in a gas-medium apparatus at a confining pressure of 300 MPa and temperatures from 1100–1200℃ on dried and annealed Aheim dunite with a grain size of ~ 400 μm. In each experiment, we performed two load relaxations. The first relaxation was initiated after rapidly loading our annealed samples to a differential stress of ~ 200 MPa within 60 s, and the second relaxation was initiated after steady-state creep was reached at a similar, constant stress. 

During the first relaxation, we find that viscosities are initially 1–2 orders of magnitude lower than steady-state viscosities before converging to the steady-state creep flow law over the course of minutes to hours. Meanwhile, such an interval of transient rheological behavior is absent during load relaxations from steady state creep. Microstructural analysis of our starting materials and deformed samples indicates that the observed transient behavior cannot be attributed to changes in grain size or crystallographic preferred orientation. Instead, the transient behavior likely corresponds to changes in dislocation density, which systematically increased during deformation following a piezometric relationship.

We compare these observations to numerical predictions of the backstress model, taking into account the stress history preceding the relaxations, the grain size and the initial dislocation density of our samples. We find that the backstress model accurately predicts the viscosity reduction during the interval of transient rheological behavior, although it slightly underestimates the duration of the transient. In addition, the absence of transient behavior during relaxation subsequent to steady-state creep indicates that the magnitude of backstress during steady-state creep is similar to the applied stress, in agreement with the model. However, the backstress model tends to overestimate strain rates during steady-state creep and subsequent relaxation. Analysis of decorated dislocations in our deformed samples indicates that this discrepancy may be due to the overestimation of dislocation density during steady-state creep by the backstress model. We discuss potential modifications to improve the model involving the effects of temperature and internal stress heterogeneity on the transient behavior of olivine.